{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:JW775INETUZE5P4N726IS4JV4D","short_pith_number":"pith:JW775INE","schema_version":"1.0","canonical_sha256":"4dbffea1a49d324ebf8dfebc897135e0d78a64f3643bcb5e741dbc2b2934f20d","source":{"kind":"arxiv","id":"2602.05435","version":2},"attestation_state":"computed","paper":{"title":"Stable Velocity: A Variance Perspective on Flow Matching","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Donglin Yang, Liang Hou, Pengfei Wan, Renjie Liao, Xiaojuan Qi, Xin Tao, Xin Yu, Yongxing Zhang","submitted_at":"2026-02-05T08:25:05Z","abstract_excerpt":"While flow matching is elegant, its reliance on single-sample conditional velocities leads to high-variance training targets that destabilize optimization and slow convergence. By explicitly characterizing this variance, we identify 1) a high-variance regime near the prior, where optimization is challenging, and 2) a low-variance regime near the data distribution, where conditional and marginal velocities nearly coincide. Leveraging this insight, we propose Stable Velocity, a unified framework that improves both training and sampling. For training, we introduce Stable Velocity Matching (Stable"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2602.05435","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-02-05T08:25:05Z","cross_cats_sorted":[],"title_canon_sha256":"2149414a7663e42ebefd2efdc80c178f7572feede933fb7f85bd72ab90982872","abstract_canon_sha256":"8d974a1785ea1d9b48996aabbdee5611ff46b5c920ab677c6591bf1426c1b436"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T02:04:50.938093Z","signature_b64":"aeNv9SjqfqgIE3lWOQYCidBX+EZk9Iw8UK85rFM2ntLrHxP8DksnQQYimbQue9KQwsyrOgn0yuSJXrlJsyD9Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4dbffea1a49d324ebf8dfebc897135e0d78a64f3643bcb5e741dbc2b2934f20d","last_reissued_at":"2026-06-02T02:04:50.937605Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T02:04:50.937605Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Stable Velocity: A Variance Perspective on Flow Matching","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Donglin Yang, Liang Hou, Pengfei Wan, Renjie Liao, Xiaojuan Qi, Xin Tao, Xin Yu, Yongxing Zhang","submitted_at":"2026-02-05T08:25:05Z","abstract_excerpt":"While flow matching is elegant, its reliance on single-sample conditional velocities leads to high-variance training targets that destabilize optimization and slow convergence. By explicitly characterizing this variance, we identify 1) a high-variance regime near the prior, where optimization is challenging, and 2) a low-variance regime near the data distribution, where conditional and marginal velocities nearly coincide. Leveraging this insight, we propose Stable Velocity, a unified framework that improves both training and sampling. For training, we introduce Stable Velocity Matching (Stable"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.05435","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2602.05435/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2602.05435","created_at":"2026-06-02T02:04:50.937663+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.05435v2","created_at":"2026-06-02T02:04:50.937663+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.05435","created_at":"2026-06-02T02:04:50.937663+00:00"},{"alias_kind":"pith_short_12","alias_value":"JW775INETUZE","created_at":"2026-06-02T02:04:50.937663+00:00"},{"alias_kind":"pith_short_16","alias_value":"JW775INETUZE5P4N","created_at":"2026-06-02T02:04:50.937663+00:00"},{"alias_kind":"pith_short_8","alias_value":"JW775INE","created_at":"2026-06-02T02:04:50.937663+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2605.21466","citing_title":"StreamGVE: Training-Free Video Editing via Few-Step Streaming Video Generation","ref_index":85,"is_internal_anchor":true},{"citing_arxiv_id":"2605.01581","citing_title":"Hyper-DP3: Frequency-Aware Right-Sizing of 3D Diffusion Policies for Visuomotor Control","ref_index":36,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JW775INETUZE5P4N726IS4JV4D","json":"https://pith.science/pith/JW775INETUZE5P4N726IS4JV4D.json","graph_json":"https://pith.science/api/pith-number/JW775INETUZE5P4N726IS4JV4D/graph.json","events_json":"https://pith.science/api/pith-number/JW775INETUZE5P4N726IS4JV4D/events.json","paper":"https://pith.science/paper/JW775INE"},"agent_actions":{"view_html":"https://pith.science/pith/JW775INETUZE5P4N726IS4JV4D","download_json":"https://pith.science/pith/JW775INETUZE5P4N726IS4JV4D.json","view_paper":"https://pith.science/paper/JW775INE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.05435&json=true","fetch_graph":"https://pith.science/api/pith-number/JW775INETUZE5P4N726IS4JV4D/graph.json","fetch_events":"https://pith.science/api/pith-number/JW775INETUZE5P4N726IS4JV4D/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JW775INETUZE5P4N726IS4JV4D/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JW775INETUZE5P4N726IS4JV4D/action/storage_attestation","attest_author":"https://pith.science/pith/JW775INETUZE5P4N726IS4JV4D/action/author_attestation","sign_citation":"https://pith.science/pith/JW775INETUZE5P4N726IS4JV4D/action/citation_signature","submit_replication":"https://pith.science/pith/JW775INETUZE5P4N726IS4JV4D/action/replication_record"}},"created_at":"2026-06-02T02:04:50.937663+00:00","updated_at":"2026-06-02T02:04:50.937663+00:00"}